[mjn]

New contributions to established articles should often be discussed on the talk page because of (a) though, and the fact that people sometimes just come in and start making major changes without looking at the work that's already been done is one thing that actively drives experts away. After a few history experts discuss the conflicting sources and an article reaches a semi-stable state after considerable discussion and additional research, someone will come in and make major changes to the article without reading any of that, usually based on an incomplete-at-best understanding of the historical sources, and sometimes reintroducing popular myths that had been previously removed after library research turned up good sources on the subject. Sometimes there are legitimate objections and an article should be changed. But also, sometimes, certain questions or disputes or uncertainties have already been discussed in previous versions of the article (often by historians who work in this field!), and it's worth understanding the work that's already been done to get the article to its present state, before rehashing the same issues.

A number of historians have given up on Wikipedia because it's too easy for people to edit articles without discussion. An expert will spend a bunch of time working out a solid, cited article that integrates the best current understanding, often in discussion with other people, then come back 6 months later and find their work has basically bitrotted. In which case, why bother doing the hard work of hashing out a good article in the first place? Some academics therefore prefer a much more bureaucratic encyclopedia model, where all changes must first be proposed and then vetted by an expert in the subject who "owns" the article; Citizendium and Scholarpedia are two attempts to build that model of encyclopedia.

[kstenerud]

In that case, would it not make sense to have metadata associated with a particular section, detailing the final consensus and how it was reached? Kind of like a comment describing a non-obvious section of code. Otherwise you have the equivalent of telling someone "Before you modify this code, look through the entire git history to see why it's the way it is today".

[mjn]

Yeah, there have been some discussions of that, but nobody has put together a MediaWiki plugin to do it afaik. Better annotation of regions in general is a longstanding wishlist, so discussions or even citations (or requests for citations) could be attached to a region of text. But then there are problems like how to handle annotations and arbitrary editing/splitting/moving of regions.

As a low-tech solution, you can actually put HTML comments in the source code, which people occasionally do, but it doesn't seem to be a very well-known option. <!-- Do NOT change birth date to XX/XX/XXXX, see talk page --> kind of things sometimes appear in the source. I haven't seen them used for longer discussions, though, just one-line "hey, watch out before you do X" things.

On contentious articles with a long history of debate people will sometimes write a summary on the talk page, so you don't have to read through the whole history of the discussion. On most articles, though, the talk page isn't huge, so I find it easy to glance at before making changes.

[sesqu]

If you have a problem with experts disagreeing about facts, arguing over them, reaching a consensus, and then other "experts" starting up the argument again - it really sounds like you're doing a poor job of communicating the researched consensus in the relevant article.

When that sort of stuff happens, I think you should make sure the incorrect information is referenced in the article as a misconception, or just have the correct information associated with citations. False information should only be erased if it is not "in the wild", so to speak.

[quanticle]

And they're both oh so successful, right?

Sure, Wikipedia isn't perfect. But I would question any method of evaluation that ends up preferring Citizendium to it.